The pattern of immunoreactivity on Western blots demonstrated by the synaptic marker synaptophysin in human PFC generally agrees with previous studies of developmental changes in synaptic density in the primate cerebral cortex (Huttenlocher 1979
; Rakic et al. 1986
; Zecevic et al. 1989
; Bourgeois and Rakic 1993
). The present data suggests robust synaptogenesis during early childhood, with synaptic density reaching a peak in late childhood, followed by the onset of synaptic elimination (a.k.a. synaptic pruning) around early adolescence, to reach adult levels by late adolescence. In his 1979 paper, Huttenlocher concluded that synaptic density peaked during the first year of life in the human and that after the age of 1 year, synaptic density persisted and declined until adolescence. However, our data show that synaptic density increases much later into childhood, from birth until late childhood (age 10) and then decreases throughout adolescence to reach the adult levels in the later teenage years. Our study therefore suggests that synaptic density increases progressively throughout childhood, instead of only during the first several years of life, as previously thought (Huttenlocher 1979
). Synaptophysin content as shown by immunohistochemistry confirmed the increasing pattern of synaptic density in early childhood but it suggested a more stable pattern of immunoreactivity during adolescence and into adulthood. The individual layers of the PFC exhibited comparable levels of synaptophysin immunoreactivity, a pattern similar to that seen in a study of the human neocortex (Masliah et al. 1990
), and which is consistent with previous studies which have shown that synaptic density is fairly equivalent across all layers (Zecevic et al. 1989
). However, two other studies (Bourgeois et al. 1994
; Huttenlocher and Dabholkar 1997
) have shown that layers 2 and 3 have a higher synaptic density than layers 1 and 4-6. A more subtle pattern of immunoreactivity between layers seen here in the immunohistochemical experiments could be due to several factors including reduced sensitivity of immunohistochemistry compared to Western blot and the fact that immunohistochemistry is inherently a less quantitative technique. Also, sectioned PFC was only available from a subset of subjects included in the Western blot experiment (25 of 42 subjects). It is interesting that the levels of synaptophysin seen via Western blotting and immunohistochemistry were similar during early development when synaptogenesis is occurring while the levels of synaptophysin differed between the two experimental methodologies mainly in the late childhood and adolescent time periods. Western blotting involves grinding up the tissue and potentially allowing better access to the antigen by the antibody while the antibody may not be able to access the antigen as well with immunohistochemistry. This could reflect that synaptophysin levels are changing within synapses at a greater level after early childhood than can readily be detected via immunohistochemistry due to the spatial constraints of the visualization method.
PSD-95 levels also exhibited an increase from birth through late childhood by Western blot. However, unlike synaptophysin, PSD-95 levels do not exhibit a significant decrease in late adolescence. These results are similar to that seen in the mouse and rodent hippocampus and cortex with Western blotting (Sans et al. 2000
) and electron microscopy (Liu et al. 2004
). Because only about 60% of excitatory synapses express PSD-95 (Aoki et al. 2001
), it is possible that those excitatory synapses which do not express PSD-95 are preferentially eliminated during adolescence. However, a recent study of cultured neurons overexpressing PSD-95 suggests that PSD-95 contributes to the synaptic maturation and stabilization of excitatory synapses (El-Husseini et al. 2000
). Therefore, increased levels of PSD-95 in adolescence and adulthood may reflect intrasynaptic maturational/stability factors rather than a continued increase in synaptic number, considered further below.
The simplest explanation of these results is that they reflect the pattern of synaptic development across perinatal and postnatal cortical development. Because we see somewhat similar patterns (i.e., increasing levels) with both synapse-associated proteins, particularly during the first ten years of postnatal development, this data suggests that changes are due to a structure common to both proteins, most likely the synapse, during early and late childhood. However, there are other explanations that may also contribute to these results. The changes observed across development, especially during adolescence when the patterns diverge between the two proteins (synaptophysin decreasing and PSD-95 plateauing), may be due to specific developmental changes in each protein occurring independently of synaptic development. For example, if the synaptic concentration of PSD-95 increases with age, this could account for the absence of a significant reduction in late adolescence. Recent data has shown that PSD-95 diffuses rapidly between synapses and PSD-95 levels determine synaptic size and strength (Gray et al. 2006
) In addition, PSD-95 is retained at synapses longer with increasing age (Gray et al. 2006
). This data, taken together with our results, suggest that the pattern of PSD-95 that we observe during adolescence could reflect synapses becoming more stable and PSD-95 being retained longer by synapses. Alternatively, the number of synaptic vesicles per synapse and/or the size of the postsynaptic density may be changing across development to further explain the pattern of synaptophysin and PSD-95 immunoreactivity, respectively. A study by Nakamura et al. (Nakamura et al. 1999
) found that early in postnatal development, the number of synaptic vesicles per synapse matures somewhat earlier than the number of synapses, suggesting that early levels of synaptophysin might reflect a greater density of synaptic vesicles per synapse than represent an accurate reflection of synapses. In addition, levels of synaptic proteins may in part reflect synaptic activity (Eastwood and Harrison 2001
). In addition, the pattern of immunoreactivity exhibited by synaptophysin parallels the pattern observed for cerebral metabolism in human cerebral cortex from birth to late adolescence using positron emission tomography studies (Chugani 1998
). This is to be expected, as glucose metabolism is thought to be an indirect measure of the number of synapses in the cortex and levels of synaptophysin have been shown to correlate with glucose metabolism (Rocher et al. 2003
). Finally, the decrease in synapse density late in adolescence corresponds well with the increasing ability to perform certain prefrontal cortical-dependent tasks (Casey et al. 2000
), consistent with the idea that normal synaptic pruning serves to improve the efficiency of synaptic connectivity.
A previous study in the human suggested that supragranular pyramidal neuron spine density, a measure of excitatory synapses, continues to decrease beyond age 20 and does not plateau at the adult level until much later in life, around age 40, at least in prefrontal area 10 and occipital area 18 (Jacobs et al. 1997
). This would suggest that if we were to continue to examine synaptic marker levels in adults over age 25, their levels would continue to decrease until the plateau corresponding to adult levels is attained. Indeed, Masliah et al (Masliah et al. 1993
) showed that synaptophysin levels did in fact continue to drop off slowly after age 40 and then decrease even more later in life.
Several potentially confounding variables in the current study were considered. First, each age group had a relatively small number of subjects which likely increased variability across groups and limited the statistical power of the study. Variability in synaptic number may particularly emerge in the pre-adolescent and adolescent groups given recent data indicating that IQ correlates with the trajectory of gray matter changes on MRI during adolescence (Shaw et al. 2006
). While unavoidable in a postmortem analysis, the cross-sectional nature of the study limits the ability to label the findings as changes across age because of individual variability (see (Kraemer et al. 2000
). The impact of postmortem stability was also considered. In rat cortex, previous studies have shown that synaptophysin and PSD-95 levels do not exhibit significant decreases with PMI less than 72 hours (Halim et al. 2003
; Hilbig et al. 2004
; Siew et al. 2004
). In human cortex, synaptophysin levels appear to be fairly robust over PMIs of less than 84 hours (but see Vawter et al. 2002
) (Honer et al. 1992
; Eastwood et al. 1994
; Eastwood et al. 1994
; Vawter et al. 2002
). These findings are consistent with our study of the postmortem stability of synaptophysin in rat cortex over 24 hours. However, in the human, PSD-95 showed a significant decrease in protein levels with postmortem intervals greater than 24 hours (Siew et al. 2004
) as was also suggested by our PMI study in rat cortex. We included PMI in our statistical model and found that PMI did not influence synaptophysin levels, but did affect those of PSD-95, in the human.
The current study provides evidence from synaptophysin and PSD-95 in human PFC that robust synaptogenesis occurs during the first ten years of life. Subsequent reductions in synaptophysin indicate that a period of synaptic elimination occurs in adolescence, particularly during the 11-15 year age range, and continues until the twenties. This developmental pattern of synaptic maturation has been proposed to allow for recovery and/or adaptation of the normal brain (Webb et al. 2001
) and for the maturation of cognitive function (Goldman-Rakic 1987
). These data shed important light on neurodevelopmental disorders such as schizophrenia, in which the age of symptom onset overlaps with the age interval during which late developmental refinement of synaptic connectivity is occurring. In fact, the importance of altered synaptic elimination in the pathogenesis of schizophrenia has long been posited (Feinberg 1982-83
) but until recently, data to support even normal developmental synaptic elimination has been scarce. Identifying the pattern of normal synaptic development of the human cortex is critical to a better understanding of the many diseases thought to be related to synaptic maldevelopment.